Research on the nasal parameters of speaker recognition system

In this paper,based on the pole-zero model of the nasal sound,relevant parameters have been extracted from the individual’s nasal voice by using linear prediction,linear prediction distortion and homomorphism theories.Ex-perimental results show that the nasal parameters are more effective than some simple frequency parameters(such as formants) in distinguishing different speakers,which means that the nasal parameters has potential value in speaker identification.